How the fate of the US economy rests on a Dell workstation
Quick, someone send Bernanke a supercomputer
The global economy is teetering on the brink of disaster.
So, you might think the quants and smart-alecs of the financial industry would have a giant supercomputer humming away somewhere, processing zillions of what-if scenarios per second and sucking in all manner of real-time economic data from the global markets and government bureaucracies to crunch some numbers and plot a logical set of economic policies to clean up this mess they got us into. No?
After hearing a report on National Public Radio's Morning Edition program describing a computerised economic model created by one-time Princeton University economic professor and present Federal Reserve chairman Ben Bernanke, called the Financial Accelerator, I certainly got the impression that the United States government had some pretty hefty simulations at its disposal to try to figure out what happened to the economy and how to fix it.
Bernanke, in case you have been hiding under a rock (a sensible course of action, really), is an expert on the economic and political forces that caused the Great Depression.
Sim Wall St
Given his expertise and the fact that Bernanke and his then colleague, New York University economics professor Mark Gertler, came up with the equations way back in the 1980s to simulate portions of the US economy as it deals with a credit crunch, you would think we would have a much more sophisticated set of tools at the disposal of the Federal Reserve. I certainly got that impression from the NPR report, which you can take a gander at here in its text version.
You might think that this massive $700bn bailout of the US financial services industry at least had some simulations backing up its numbers - particularly because we live in a world where we can view the weather live from satellites, and gather an enormous amount of physical data about air and water and land that can be put into simulations that can tell us with at least some accuracy what the weather will be like anywhere on the planet within a few hours.
Uncle Sam has dedicated several petaflops of computing power to redesigning nuclear war heads and simulating nuclear explosions (thereby getting around the Nuclear Test Ban Treaty, which forbids actually blowing up a real new nuke design to test it). You'd think at least one central bank (most likely the Federal Reserve Bank, given the central nature of the US economy since the end of World War II) would have a supercomputer cranking through scenarios right now. After all, the big banks and brokerages are doing Monte Carlo simulations to assess the risks of buying and selling securities using more or less real-time stock data.
This seems like a reasonable assumption, right? Right.
So I contacted Gertler to try to get some kind of understanding about how the Financial Accelerator model works, and what supercomputer it is implemented on. The acceleration that the model is talking about is how a credit crunch, caused by a crisis in confidence in some aspect of the economy, builds on itself and accelerates a decline. (You can read a 1996 paper put together by Bernanke, Gertler and their partner, professor Simon Gilchrist of Boston University, called The Financial Accelerator and the Flight to Quality here (pdf), and an interesting follow-up speech by Bernanke given in June 2007 called The Financial Accelerator and the Credit Channel here (pdf).)
Like the rest of you, I wanted to know what the model was saying giving the current conditions in the US stock, credit, and jobs markets.
I was able to reach Gertler by email. "Things are pretty hectic for me this week," he explained. An understatement of comic proportions had this not been such a serious bit of business.
When I asked how the Financial Accelerator was implemented, Gertler said: "The model is a set of non-linear equations. We take a loglinear approximation around the steady-state, which gives us a set of difference equations. We then solve use software such as Matlab or Dynare."
In the NPR piece, Gertler said that the Financial Accelerator model had been used in the South Korean economic crisis a decade ago, where gross domestic product for that country dropped by 12 per cent - with half of that drop coming due to the effects of feedback loops of ever-tightening credit.
Gertler added that his colleague at Boston University was using the model to look at current events, so I contacted him.
You wanted a bailout? You got a Dell dude
As it turns out, Gilchrist's version of the Financial Accelerator model is running on a Dell workstation, and it implements the 20 equations to describe the economy during a credit crunch in a programming language called Matlab from MathWorks. Graduate students run the model, which takes GDP, economic output, investment spending, employment figures, corporate bond spreads, and other economic data to figure out how these changing figures will affect the economy.
A simulation run to solve the equations takes a few hours. "Even before the September meltdown, the pressures in the market have been reducing GDP by two per cent," Gilchrist says. This is due to the financial accelerator effect, adding that this has been partially offset by relatively loose monetary policy.
Interestingly, the Financial Accelerator simulation does not model the financial services sector, and therefore Bernanke and secretary of the Treasury Hank Paulson have not been calling Gilchrist up to run some numbers.
According to Gilchrist, the Bank of England and the Bank of Canada have put together some computerised simulations of their countries' respective economies, and the Fed was talking about possibly putting together a more sophisticated model. But there are limits, apparently.
"These types of models are rarely used for economic forecasts," explains Gilchrist.
But why not? Why isn't there a supercomputer somewhere, right now, simulating the effects of a $700bn bailout?
I think right now we'd all settle for a real good spreadsheet, and maybe the thing to do is get Gilchrist a few more grad students, a baby supercomputer like he says he will need to be able to pump real-time data into the model, and get us to a point where we can do some more sophisticated "what-if" modelling on economic policy. Put an earmark for the project right into the bill as it passes from the House to the Senate.
I studied rocket science, and I can tell you - this ain't rocket science.®